Transdiagnostic compulsivity is associated with reduced reminder setting, only partially attributable to overconfidence

  1. Annika Boldt  Is a corresponding author
  2. Celine Ann Fox
  3. Claire M Gillan
  4. Sam Gilbert
  1. Institute of Cognitive Neuroscience, University College London, United Kingdom
  2. School of Psychology, Trinity College Dublin, Ireland
38 figures, 15 tables and 5 additional files

Figures

Average accuracy as a function of whether a reminder was used.

‘No Reminder’: forced internal condition; ‘Reminder’: forced external condition. Each pair of dots linked by a line indicates one participant. The red data points are excluded participants. The box plots indicate the median surrounded by the interquartile range (25th and 75th percentile). The whiskers show the minimum and maximum. The preregistered exclusion criteria for the accuracies with or without reminder are indicated as horizontal dotted lines (10% and 70%, respectively).

People’s tendency to set reminders above or below the optimal offloading strategy (reminder bias) plotted against people’s tendency towards over- or underconfidence (metacognitive bias).

The solid line indicates the fitted relationship between both variables. The dashed lines represent the 95% confidence interval around it. Each circle represents a single participant.

Standardised regression weights for the ‘anxious-depression’ (AD) factor and the ‘compulsive behaviour and intrusive thought’ (CIT) factor predicting metacognitive bias.

Error bars indicate 95% confidence intervals. Asterisks indicate significance: ‘***’: <0.001; ‘**’: <0.01; ‘*’: <0.05.

Standardised regression weights for the ‘anxious-depression’ (AD) factor and the ‘compulsive behaviour and intrusive thought’ (CIT) factor predicting reminder bias.

Error bars indicate 95% confidence intervals. Asterisks indicate significance: ‘***’: <0.001; ‘**’: <0.01; ‘*’: <0.05.

Diagram of the mediation analysis testing for the influence of the ‘compulsive behaviour and intrusive thought’ (CIT) factor on reminder bias, both directly and indirectly through the metacognitive bias.

Standardised regression coefficients are given for each path. The value in parentheses indicates the influence of CIT on reminder bias controlling for the influence of the metacognitive bias. Asterisks indicate significance: ‘***’: <0.001; ‘**’: <0.01; ‘*’: <0.05.

Overview of the intention offloading paradigm.

(A) Example sequence of events within a single trial. Trajectories of movement made by a fictive participant are shown as black arrows. The blue coloured circle corresponds to the left boundary of the square and indicates that this circle must be moved to this side rather than the bottom. (B) Example of an offloading decision which participants were required to make before each trial. (C) After each decision, they were informed whether or not they would perform the upcoming trial with reminders. The cell’s shading indicates the participant’s original choice. (D) Confidence was rated once before the introduction of the offloading strategy on a scale ranging from 0% to 100%. (E) Sequence of events within the task. All aspects of the task were performed online in the web browser.

Appendix 1—figure 1
The actual indifference point (AIP) is shown on the x-axis against its recovered estimates on the y-axis.

Each marker represents one participant’s estimates.

Appendix 1—figure 2
Psychometric functions linking target values to offloading choices.

The average choice data is shown as dots. Panels show the individual curves for participants 1–20.

Appendix 1—figure 3
Psychometric functions linking target values to offloading choices.

The average choice data is shown as dots. Panels show the individual curves for participants 21–40.

Appendix 1—figure 4
Psychometric functions linking target values to offloading choices.

The average choice data is shown as dots. Panels show the individual curves for participants 41–60.

Appendix 1—figure 5
Psychometric functions linking target values to offloading choices.

The average choice data is shown as dots. Panels show the individual curves for participants 61–80.

Appendix 1—figure 6
Psychometric functions linking target values to offloading choices.

The average choice data is shown as dots. Panels show the individual curves for participants 81–100.

Appendix 1—figure 7
Psychometric functions linking target values to offloading choices.

The average choice data is shown as dots. Panels show the individual curves for participants 101–120.

Appendix 1—figure 8
Psychometric functions linking target values to offloading choices.

The average choice data is shown as dots. Panels show the individual curves for participants 121–140.

Appendix 1—figure 9
Psychometric functions linking target values to offloading choices.

The average choice data is shown as dots. Panels show the individual curves for participants 141–160.

Appendix 1—figure 10
Psychometric functions linking target values to offloading choices.

The average choice data is shown as dots. Panels show the individual curves for participants 161–180.

Appendix 1—figure 11
Psychometric functions linking target values to offloading choices.

The average choice data is shown as dots. Panels show the individual curves for participants 181–200.

Appendix 1—figure 12
Psychometric functions linking target values to offloading choices.

The average choice data is shown as dots. Panels show the individual curves for participants 201–220.

Appendix 1—figure 13
Psychometric functions linking target values to offloading choices.

The average choice data is shown as dots. Panels show the individual curves for participants 221–240.

Appendix 1—figure 14
Psychometric functions linking target values to offloading choices.

The average choice data is shown as dots. Panels show the individual curves for participants 241–260.

Appendix 1—figure 15
Psychometric functions linking target values to offloading choices.

The average choice data is shown as dots. Panels show the individual curves for participants 261–280.

Appendix 1—figure 16
Psychometric functions linking target values to offloading choices.

The average choice data is shown as dots. Panels show the individual curves for participants 281–300.

Appendix 1—figure 17
Psychometric functions linking target values to offloading choices.

The average choice data is shown as dots. Panels show the individual curves for participants 301–320.

Appendix 1—figure 18
Psychometric functions linking target values to offloading choices.

The average choice data is shown as dots. Panels show the individual curves for participants 321–340.

Appendix 1—figure 19
Psychometric functions linking target values to offloading choices.

The average choice data is shown as dots. Panels show the individual curves for participants 341–360.

Appendix 1—figure 20
Psychometric functions linking target values to offloading choices.

The average choice data is shown as dots. Panels show the individual curves for participants 361–380.

Appendix 1—figure 21
Psychometric functions linking target values to offloading choices.

The average choice data is shown as dots. Panels show the individual curves for participants 381–400.

Appendix 1—figure 22
Psychometric functions linking target values to offloading choices.

The average choice data is shown as dots. Panels show the individual curves for participants 401–420.

Appendix 1—figure 23
Psychometric functions linking target values to offloading choices.

The average choice data is shown as dots. Panels show the individual curves for participants 421–440.

Appendix 1—figure 24
Psychometric functions linking target values to offloading choices.

The average choice data is shown as dots. Panels show the individual curves for participants 441–460.

Appendix 1—figure 25
Psychometric functions linking target values to offloading choices.

The average choice data is shown as dots. Panels show the individual curves for participants 461–480.

Appendix 1—figure 26
Psychometric functions linking target values to offloading choices.

The average choice data is shown as dots. Panels show the individual curves for participants 481–500.

Appendix 1—figure 27
Psychometric functions linking target values to offloading choices.

The average choice data is shown as dots. Panels show the individual curves for participants 501–520.

Appendix 1—figure 28
Psychometric functions linking target values to offloading choices.

The average choice data is shown as dots. Panels show the individual curves for participants 521–540.

Appendix 1—figure 29
Psychometric functions linking target values to offloading choices.

The average choice data is shown as dots. Panels show the individual curves for participants 541–560.

Appendix 1—figure 30
Psychometric functions linking target values to offloading choices.

The average choice data is shown as dots. Panels show the individual curves for participants 561–580.

Appendix 1—figure 31
Psychometric functions linking target values to offloading choices.

The average choice data is shown as dots. Panels show the individual curves for participants 581–600.

Appendix 1—figure 32
The distribution of the unaided task performance (internal accuracy) as a function of ‘compulsive behaviour and intrusive thought’ (CIT; left panel) and ‘anxious depression’ (AD; right panel).

Tables

Key resources table
Reagent type (species) or resourceDesignationSource or referenceIdentifiersAdditional information
Software, algorithmRR Development Core Team, 20244.4.2; RRID:SCR_001905
Software, algorithmRStudioRStudio Team, 20202024.09.1+394; RRID:SCR_000432
Software, algorithmdiagramSoetaert, 20201.6.5; RRID:SCR_026982R package
Software, algorithmeffectsizeBen-Shachar et al., 20200.8.9; RRID:SCR_026983R package
Software, algorithmlmerTestKuznetsova et al., 20173.1-3; RRID:SCR_015656R package
Software, algorithmlme4Bates et al., 20151.1-35.5; RRID:SCR_015654R package
Software, algorithmmediationTingley et al., 20144.5.0; RRID:SCR_026984R package
Software, algorithmplyrWickham, 20111.8.9; RRID:SCR_026985R package
Software, algorithmpwrChampely, 20201.3-0; RRID:SCR_025480R package
Software, algorithmquickpsyLinares and López-Moliner, 20160.1.5.1; RRID:SCR_026986R package
Table 1
List of preregistered hypotheses together with the empirical support our study found.

White background indicates sanity check hypotheses, and grey background indicates key hypotheses. OIP = optimal indifference point. AIP = actual indifference point. CIT = compulsive behaviour and intrusive thought.

NumberHypothesisSupport?
H1The reminder bias and metacognitive bias are negatively correlated.Yes
H2Participants use reminders excessively.Yes
H3Participants are underconfident in their own memory.Yes
H4OIP and AIP are positively correlated.Yes
H5aPositive link between CIT and reminder bias.No (significant negative effect)
H5bPositive link between CIT and absolute number of reminders chosen.No (negative effect but significance not reached)
H5cPositive link between CIT and AIP.No (significant negative effect)
H5dPositive link between CIT and reminder bias even if working memory is included as a covariate.No (significant negative effect)
H5ePositive link between CIT and reminder bias even if cognitive ability is included as a covariate.No (significant negative effect)
H6aA significant link exists between CIT and metacognitive bias (preregistered as a two-sided test, so either more or less confident).Yes (positive)
H6bA significant link exists between CIT and raw confidence.Yes (positive)
H6cA significant link exists between CIT and metacognitive bias even if cognitive ability is included as a covariate.Yes (positive)
H7aCIT acts as a moderator on the link between confidence and offloading. In other words, we expect to find that the correlation between the metacognitive and the reminder bias to be weakened in highly compulsive individuals.No
H7bCIT acts as a moderator on the link between confidence and offloading even if working memory is included as a covariate.No
H7cCIT acts as a moderator on the link between confidence and offloading even if cognitive ability is included as a covariate.No
H8aA significant negative link exists between AD and metacognitive bias (i.e. more anxious-depressed individuals tend to be underconfident).Yes
H8bA significant negative link exists between AD and raw confidence.Yes
H8cA significant negative link exists between AD and metacognitive bias even if cognitive ability is included as a covariate.Yes
Appendix 1—table 1
Predicting metacognitive bias.

All continuous variables are z-transformed. SE = standard error; m=male; f=female; o=other.

βSEtp
Intercept0.070.051.300.193
AD–0.230.05–4.99<0.001
CIT0.150.053.110.002
Age–0.020.04–0.550.586
gender1 (m vs. f)–0.170.08–2.060.040
gender2 (m vs. o)–0.290.38–0.780.438
education–0.00010.04–0.0050.996
Appendix 1—table 2
Predicting confidence.

All continuous variables are z-transformed. SE = standard error; m=male; f=female; o=other.

βSEtp
Intercept0.090.051.810.070
AD–0.290.04–6.43<0.001
CIT0.120.052.760.006
Age–0.140.04–3.44<0.001
gender1 (m vs. f)–0.240.08–2.930.004
gender2 (m vs. o)–0.230.37–0.630.528
education0.040.041.010.311
Appendix 1—table 3
Predicting metacognitive bias with ICAR5 scores as an additional covariate.

All continuous variables are z-transformed. SE = standard error; m=male; f=female; o=other.

βSEtp
Intercept0.080.051.680.094
AD–0.200.05–4.46<0.001
CIT0.120.052.570.011
Age–0.030.04–0.660.507
gender1 (m vs. f)–0.220.08–2.610.009
gender2 (m vs. o)–0.450.37–1.210.226
education0.040.040.910.364
ICAR5–0.200.04–4.84<0.001
Appendix 1—table 4
Predicting reminder bias.

All continuous variables are z-transformed. SE = standard error; m=male; f=female; o=other.

βSEtp
Intercept–0.010.05–0.240.813
AD0.070.051.460.146
CIT–0.140.05–2.910.004
Age0.070.041.690.092
gender1 (m vs. f)0.0050.080.060.955
gender2 (m vs. o)0.880.382.320.021
education–0.060.04–1.420.157
Appendix 1—table 5
Predicting absolute number of reminders.

All continuous variables are z-transformed. SE = standard error; m=male; f=female; o=other.

βSEtp
Intercept–0.030.05–0.680.496
AD0.060.051.330.183
CIT–0.090.05–1.940.053
Age0.180.044.38<0.001
gender1 (m vs. f)0.070.080.860.393
gender2 (m vs. o)0.730.381.930.054
education–0.100.04–2.580.010
Appendix 1—table 6
Predicting actual indifference point (AIP).

All continuous variables are z-transformed. SE = standard error; m=male; f=female; o=other.

βSEtp
Intercept0.020.050.450.657
AD–0.080.05–1.760.079
CIT0.100.052.250.025
Age–0.170.04–3.95<0.001
gender1 (m vs. f)–0.040.08–0.450.652
gender2 (m vs. o)–0.750.38–1.990.047
education0.090.042.240.025
Appendix 1—table 7
Predicting reminder bias with 2-back d’ as an additional covariate.

All continuous variables are z-transformed. SE = standard error; m=male; f=female; o=other.

βSEtp
Intercept–0.010.05–0.230.821
AD0.060.051.230.219
CIT–0.120.05–2.570.010
Age0.070.041.780.076
gender1 (m vs. f)0.0040.080.050.961
gender2 (m vs. o)0.860.382.250.025
education–0.060.04–1.560.120
2-back d’0.100.042.410.016
Appendix 1—table 8
Predicting reminder bias with ICAR5 scores as an additional covariate.

All continuous variables are z-transformed. SE = standard error; m=male; f=female; o=other.

βSEtp
Intercept–0.010.05–0.260.796
AD0.070.051.410.160
CIT–0.140.05–2.850.005
Age0.070.041.700.091
gender1 (m vs. f)0.010.080.090.927
gender2 (m vs. o)0.900.382.330.020
education–0.060.04–1.450.147
2-back d’0.010.040.320.751
Appendix 1—table 9
Predicting reminder bias with metacognitive bias as an additional covariate (i.e. testing for a moderation effect).

All continuous variables are z-transformed. SE = standard error; m=male; f=female; o=other; MetaBias = metacognitive bias.

βSEtp
Intercept–0.010.05–0.250.802
Metacognitive bias–0.190.04–4.66<0.001
AD0.030.050.670.506
CIT–0.100.05–2.140.032
Age0.130.043.220.001
gender1 (m vs. f)0.0030.080.040.969
gender2 (m vs. o)0.990.372.650.008
education–0.080.04–2.030.043
CIT X MetaBias–0.010.04–0.180.857
Appendix 1—table 10
Predicting reminder bias with metacognitive bias and 2-back d’ as additional covariates.

All continuous variables are z-transformed. SE = standard error; m=male; f=female; o=other; MetaBias = metacognitive bias.

βSEtp
Intercept–0.010.05–0.260.797
Metacognitive Bias–0.170.04–4.27<0.001
AD0.030.050.550.584
CIT–0.090.05–1.900.058
Age0.140.043.280.001
gender1 (m vs. f)0.0040.080.060.953
gender2 (m vs. o)0.970.372.600.010
education–0.090.04–2.130.034
2-back d’0.080.041.950.052
CIT X MetaBias–0.010.04–0.260.793
Appendix 1—table 11
Predicting reminder bias with metacognitive bias and ICAR5 scores as additional covariates.

All continuous variables are z-transformed. SE = standard error; m=male; f=female; o=other; MetaBias = metacognitive bias.

βSEtp
Intercept–0.010.05–0.270.789
Metacognitive bias–0.190.04–4.57<0.001
AD0.030.050.640.521
CIT–0.100.05–2.110.035
Age0.130.043.220.001
gender1 (m vs. f)0.0050.080.060.949
gender2 (m vs. o)1.000.372.660.008
education–0.080.04–2.030.043
ICAR50.0090.040.220.829
CIT X MetaBias–0.010.04–0.180.859
Appendix 1—table 12
Predicting reminder bias with metacognitive bias and ICAR5 scores as additional covariates.

All continuous variables are z-transformed. SE = standard error; m=male; f=female; o=other; MetaBias = metacognitive bias.

βSEtp
Intercept–0.020.05–0.370.712
Metacognitive bias–0.190.04–4.65<0.001
AD0.030.050.670.501
CIT–0.100.05–2.150.032
Age0.130.043.150.002
gender1 (m vs. f)0.0040.080.040.966
gender2 (m vs. o)0.990.372.650.008
education–0.080.04–1.910.057
AD X MetaBias–0.040.04–0.940.349
Appendix 1—table 13
Predicting internal accuracy.

All continuous variables are z-transformed. SE = standard error; m=male; f=female; o=other.

βSEtp
Intercept0.010.050.270.784
AD–0.020.05–0.390.693
CIT–0.060.05–1.330.183
Age–0.150.04–3.50<0.001
gender1 (m vs. f)–0.040.08–0.520.606
gender2 (m vs. o)0.150.380.400.687
education0.050.041.290.199

Additional files

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Annika Boldt
  2. Celine Ann Fox
  3. Claire M Gillan
  4. Sam Gilbert
(2025)
Transdiagnostic compulsivity is associated with reduced reminder setting, only partially attributable to overconfidence
eLife 13:RP98114.
https://doi.org/10.7554/eLife.98114.4